Field notes on AI, automation and the systems-first stack we use to run a startup with a team of two — and now apply to other businesses.
OpenAI's new image model is the best thing anyone has shipped. But ChatGPT can't post anywhere. The next year is a race for the distribution layer.
I'm Archie Roberts, co-founder of rheos.app. This is what I'm building, how I got here, and why I'm writing about it.
Why we built rheos.app, the problem it solves, and where we're at 7 months in.
Meta's new model searches live across Facebook, Instagram, WhatsApp and Threads — multimodally. If your business isn't posting consistently, you're about to become invisible.
Opus 4.7 is the first model where using tools actually works end to end, and that changes what tech companies need to optimise for.
How 20+ MCP servers close the feedback gaps that slow down small-team development — real examples from building a SaaS with Claude Code.
How one Claude Code setup powers three products — a social media platform, an investment tool, and a knowledge OS — using layered config, custom skills, MCP servers, and a shared knowledge vault.
Every layer of the tech stack behind rheos.app — and how AI tooling lets two founders ship like a team of ten.
Claude starts every conversation from zero. Skills and hooks are how I gave it persistent memory of our stack, our patterns, and our decisions.
How I turned an Obsidian vault into a searchable, embeddable knowledge system with a multi-model AI council and an MCP server for Claude.
How I built an MCP server so Claude can see my investment portfolio, and a dashboard to go with it.
Two AI apps at different stages — making professional-grade tools accessible to normal people through AI.
Meta's new model fills the real-time social media search gap that Grok has been covering. But the real play is using multiple models together — and I'm building an app around that.